In this study the material from “Five year study of Epidemiology of Tuberculosis” (1961-68) has been analysed to find out an improved method of interpretation of chest X-rays to get accurate estimation of prevalence of “suspects” in the community. The population of a random sample of 119 villages from the three taluks of Bangalore district was surveyed four times with intervals of 1½ to 2 years by tuberculin testing, 70mm chest photofluorography and sputum bacteriology. Out of 45,434 persons X-rayed during the first survey, 590 were read as active pulmonary tuberculosis on the basis of single picture interpretation by two independent readers. Of them, 460 being sputum culture negative were classified as initial “suspects” and these were reviewed in this study by the panel of three readers together by the method of “joint reading”. The interpretation was done comparing the serial X-rays of individuals taken at intervals along with other available examination results and personal data. Out of 460 initial suspects only 110 (23.9%) were confirmed as “suspects”, the remaining were judged as non-tuberculous and/or inactive tuberculous (62.2%) and normals (13.9%).

Fates on five year follow up were compared between 85 “confirmed suspects” and 385 “initial suspects”. The mortality and sputum positive status were found more among the former group i.e., 23.5 and 25.5 and 14% and 7.2% respectively. Radiologically, 48.7% of the confirmed suspects and only 10% of the initial suspects could be classified as suspects at 5th year follow up. Incidence of bacillary disease among the confirmed suspects was also found higher. On the basis of “joint reading” and five year follow up study, the limitations of single picture interpretation resulting in considerable over diagnosis were clearly seen. The comparative reading of serial X-rays along with other examination results did help in the better assessment of etiology and activity status of disease. Of the X-rays read as non-tuberculous and inactive tuberculous when reviewed by “joint reading” method, about 67 more suspects could be added. Even then the estimates of prevalence of “suspects” based on single film interpretation which are widely used in India appear to be about 3 times the actual prevalence.